DataFusion's tokio runtime panics after fork(). This is a fundamental issue for PyTorch DataLoader usage. A close()/shutdown() method on XarrayContext that cleanly tears down the runtime (so it can be recreated in child processes) would help. Or a is_valid() check.
DataFusion's tokio runtime panics after fork(). This is a fundamental issue for PyTorch DataLoader usage. A
close()/shutdown()method onXarrayContextthat cleanly tears down the runtime (so it can be recreated in child processes) would help. Or ais_valid()check.